iTRAQ- and MRM-Based Proteomics Identify Early Injury Biomarkers for Primary Dysfunction After Liver Transplantation

Introduction

Liver transplantation (LT) is the only effective treatment for end-stage liver disease. Liver grafts are classified into 3 categories based on liver function recovery: optimal graft function (OG), early allograft dysfunction (EAD), and primary nonfunction (PNF). The quality of liver allografts is closely linked to ischemia-reperfusion injury (IRI). Severe IRI can lead to EAD or PNF, both of which significantly affect patient and allograft survival [1–4].

PNF is a severe complication of LT, characterized by unexpected graft dysfunction, often requiring immediate retransplantation or resulting in patient death [5]. In Canada and France, the annual loss rate for transplanted organs can reach as high as 10% [6]. PNF is a multifactorial syndrome, with risk factors including high donor body mass index (BMI), moderate to severe steatosis, prolonged cold or warm ischemic times, and grafts from donation after circulatory death (DCD) [7]. Studies link cold ischemia times over 10 h and warm ischemia times exceeding 40 minutes with higher PNF incidence [8–10]. Additionally, DCD grafts are a significant risk factor for PNF [11]. Other donor-related factors, such as prolonged intensive care unit (ICU) stays (>5 days), hypernatremia, and diabetes, also increase PNF risk [12]. Given its complexity, no single factor or biomarker reliably predicts outcomes. Currently, the primary strategy to prevent PNF involves careful selection of donors and recipients based on known risk factors, including donor laboratory tests, clinical risk scores, organ appearance, and, in some cases, biopsies and perfusion parameters. However, these methods are limited and insufficient to fully prevent PNF [13]. The ideal biomarker for PNF should demonstrate high sensitivity for the early detection of graft injury, along with sufficient specificity to reliably distinguish PNF from other post-transplant complications that present with similar clinical features, such as acute rejection, biliary complications, or vascular insufficiency. This discriminative capacity is of critical importance, given that the therapeutic strategies are fundamentally different. For instance, PNF primarily results from severe IRI and often necessitates supportive care or retransplantation, whereas acute rejection requires immediate immunosuppressive intervention [14,15]. A biomarker that lacks specificity can result in diagnostic uncertainty, potentially delaying appropriate management or exposing patients to unnecessary and potentially ineffective treatments [16]. It is therefore vital to identify molecular signatures that uniquely reflect the pathophysiological processes underlying PNF, including mitochondrial dysfunction, microcirculatory disturbance, and hepatocyte necrosis, to clarify early differential diagnosis and inform accurate medicine strategies following LT. Advancements in technologies such as imaging, transcriptomics, genomics, proteomics, metabolomics, lipidomics, and organ perfusion have provided new insights into the mechanisms underlying IRI, inflammation, and graft rejection. These technologies offer the potential to identify biomarkers that could help predict PNF risk after LT and aid in the acceptance or rejection of suboptimal liver grafts. Such biomarker panels would likely need to account for multiple factors involved in the transplantation process.

Kornasiewicz et al applied a proteomic approach to identify protein phenotypes associated with PNF after LT, but without further validation [17]. Isobaric tags for relative and absolute quantitation (iTRAQ) is a powerful proteomic method that combines stable isotope labeling with tandem mass spectrometry for precise comparative quantification. Multiple reaction monitoring (MRM) is a targeted proteomic technique known for its high reproducibility and sensitivity in complex samples. Several studies have demonstrated the feasibility of using MRM-MS for quantitative proteomic analysis, offering multiplexed, precise, specific, and standardized quantification. Moreover, iTRAQ discovery followed by MRM verification has been used in biomarker quantification studies [18], but its application for predicting the quality of donor livers has been limited.

In this study, we used iTRAQ as a discovery tool, followed by MRM verification, to conduct a comparative proteomic analysis aimed at identifying key proteins indicative of potential hepatic injury in grafts at risk of PNF after LT. Our goal was to identify early biomarkers associated with PNF that can be measured after LT, providing insights for individualized clinical therapies to prevent PNF.

Material and Methods

PROTEIN PREPARATION AND SDS-PAGE ELECTROPHORESIS:

Proteins were extracted using lysis buffer (1 mM PMSF, 2 mM EDTA, 7 M urea, 2 M thiourea, 4% CHAPS, 40 mM Tris-HCl, pH 8.5). After 5 min, 10 mM DTT was added for reduction. The homogenized tissue was centrifuged at 25 000×g for 15 min at 4°C. The supernatant was further reduced with DTT, incubated at 56°C for 1 h, and cysteine residues were blocked with 55 mM iodoacetamide (IAM) on ice for 45 min. Proteins were precipitated with chilled acetone, incubated for 2 h, and centrifuged. The pellet was dissolved in 0.5 M TEAB, sonicated, and the supernatant was collected for analysis. Protein concentrations were determined by Bradford assay. SDS-PAGE was performed to assess protein integrity. Thirty micrograms of protein from each sample were loaded onto 10% gels, stained with Coomassie Brilliant Blue, and selected for trypsin digestion and LC-MS/MS analysis based on clear, defined bands.

QUANTITATIVE ITRAQ ANALYSIS:

One hundred micrograms of protein from each sample group were digested with 2.5 μg trypsin (Promega, USA) at a 40: 1 protein-to-trypsin ratio at 37°C for 4 h, followed by an additional 8-h incubation for complete digestion. The resulting peptides were desalted using a Strata X C18 column (Phenomenex, USA) and dried by vacuum centrifugation. Peptides were dissolved in 0.5 M TEAB and labeled with 8-plex iTRAQ reagents (Applied Biosystems, USA) according to the manufacturer’s protocol. The PNF2, PNF1, EAD2, EAD1, OG2, OG1, and Control groups were labeled with 7 distinct iTRAQ tags. Each reagent was dissolved in 24 μL isopropanol, and peptides were incubated with the reagent at room temperature for 2 h before being dried via vacuum centrifugation.

Strong Cation Exchange (SCX) chromatography was performed using an LC-20AB HPLC pump system (Shimadzu, Japan). The labeled peptides were combined, reconstituted in 2 mL buffer (5% acetonitrile [ACN], pH 9.8), and loaded onto a 4.6×250 mm Gemini C18 column (Thermo Fisher Scientific, USA). Peptides were eluted at 1 mL/min with the following gradient: 0–10 min: 5% buffer (95% ACN, pH 9.8); 10–50 min 5–35% buffer; 50–51 min: 35–90% buffer. Fractions were collected every minute and monitored at 214 nm absorbance, resulting in 20 fractions. These fractions were desalted and vacuum dried for further analysis.

Each fraction was dissolved in 2% ACN, 0.1% formic acid (FA) and centrifuged at 20 000×g for 10 min. The supernatant was loaded onto a C18 trap column (75 μm inner diameter, Phenomenex, USA) in an LC-20AD nano HPLC system (Shimadzu, Japan). Peptides were separated at a flow rate of 300 nL/min using the following gradient: 0–8 min: 5% buffer (98% ACN, 0.1% FA); 8–43 min: 5–35% buffer; 43–48 min: 35–60% buffer; 48–50 min: 60–80% buffer; 50–55 min: 80% buffer; 55–56 min: return to 5% buffer.

The separated peptides were analyzed with a Q-Exactive tandem mass spectrometer (Thermo Fisher Scientific, USA) via nanospray ionization. Data acquisition was performed in data-dependent acquisition (DDA) mode, with full MS scans obtained using the Orbitrap mass analyzer across a 350–1600 m/z range at a resolution of 70 000. The 20 most intense multiply charged precursor ions (intensity >10 000) were selected for fragmentation via higher-energy collision-induced dissociation (HCD). Tandem mass spectra were acquired in the Orbitrap at a resolution of 17 500 at 100 m/z. A 15-s dynamic exclusion was applied to prevent repeated sequencing of the same precursor ions.

DATA PROCESSING AND PROTEIN IDENTIFICATION:

All primary mass spectrometry data were converted to MGF (Mascot Generic Format) files for downstream analysis. Protein identification was performed using the Mascot search engine (Matrix Science, London, UK) integrated within the IQuant software. To ensure high-confidence identification and significantly reduce false positives, the Mascot Percolator algorithm was used. A strict false discovery rate (FDR) threshold of ≤1% was enforced at both the peptide-spectrum match (PSM) and protein levels using the Picked protein FDR strategy.

DIFFERENTIAL EXPRESSION ANALYSIS:

For the iTRAQ-based discovery phase, proteins were considered candidate differentially expressed proteins if they met the following criteria: a fold change ≥1.2, an P<0.05 (t test), and reproducibility in at least 3 independent experiments. To maximize the discovery potential from highly heterogeneous clinical liver samples and to minimize Type II errors, strict multiple testing corrections were intentionally not applied to the P values during this exploratory screening phase. This liberal threshold was chosen to capture a broad range of potentially relevant proteins for hypothesis generation, with an explicit design in which a rigorous false-positive control would be strictly enforced in the downstream targeted MRM validation phase.

FUNCTIONAL ANNOTATION AND ENRICHMENT ANALYSIS:

For functional annotation, protein sequences were analyzed using Blast2GO by comparing them with the non-redundant protein (NR) database from NCBI. The KEGG (Kyoto Encyclopedia of Genes and Genomes) database and COG (Clusters of Orthologous Groups of Proteins) database were used to classify proteins based on their biological pathways, functional roles, and orthologous relationships. Enrichment analysis was conducted to identify key metabolic and regulatory pathways associated with the differentially expressed proteins. For each GO term and KEGG pathway, enrichment significance was assessed using the hypergeometric test (or Fisher’s exact test). Consistent with the exploratory, hypothesis-generating nature of the discovery phase, terms and pathways with an P<0.05 were considered significantly enriched. Multiple testing corrections were not applied here to avoid masking broader biological trends at this preliminary stage.

MRM VALIDATION OF DIFFERENTIALLY EXPRESSED PROTEINS IDENTIFIED BY ITRAQ:

A total of 10 differentially expressed proteins were selected for MRM validation. Samples were digested with trypsin as described previously and spiked with 50 fmol β-galactosidase for label-free data normalization. MRM analysis was performed on a QTRAP 5500 mass spectrometer (SCIEX, USA) coupled with an LC-20AD nano HPLC system (Shimadzu, Japan). The mobile phase consisted of Solvent A (0.1% aqueous formic acid) and Solvent B (98% acetonitrile, 0.1% formic acid). Peptide separation was carried out on a C18 column (0.075×150 mm, 3.6 μm particle size) at 300 nL/min with the following gradient: 0–38 min: 5–30% solvent B; 38–42 min: 30–80% solvent B; and 42–50 min: 80% solvent B (maintenance). The QTRAP 5500 was operated with a spray voltage of 2400 V, nebulizer gas pressure of 23 psi, and a dwell time of 10 ms per transition. Multiple MRM transitions were monitored at unit resolution in both Q1 and Q3 quadrupoles to maximize specificity.

Raw data were processed using Skyline software, applying the iRT (indexed retention time) strategy to compare peptide spectra against a pre-established spectral library. All transitions for each peptide were used for quantification unless interference from the sample matrix was observed. The MRM signal was defined by the detection of all transitions from the endogenous peptide coeluting with the stable isotope-labeled peptide, confirmed by equivalent retention times and consistent relative areas of light and heavy transitions. Precision was evaluated by calculating the coefficient of variation (CV) as the standard deviation divided by the mean.

For the MRM validation of pre-selected candidate proteins, we applied a stringent statistical approach using MS stats with a linear mixed-effects model. An FDR correction was applied using the Benjamini–Hochberg method with a cutoff of 0.05. Proteins were considered significantly differentially expressed if the FDR <0.05 and fold change ≥1.5. The more stringent fold-change threshold was chosen to prioritize robust candidates with larger-magnitude changes and to reduce false positives.

STATISTICAL ANALYSIS:

Statistical analyses were performed using SPSS version 22.0 and GraphPad Prism version 8.0. Continuous variables were expressed as mean±standard error of the mean (SEM) or median with interquartile range (IQR), and categorical variables as frequencies and percentages. Normality was assessed using the Shapiro–Wilk test and Q-Q plots, and homogeneity of variances by Levene’s test. For comparisons between 2 groups, the t test was used for normally distributed data with homogeneous variances; otherwise, the Mann–Whitney U test was applied. For multiple-group comparisons, one-way ANOVA or the Kruskal-Wallis test was used, depending on whether parametric assumptions were met. Categorical variables were compared using the chi-square test or Fisher’s exact test. All tests were 2-tailed, and P<0.05 was considered statistically significant.

Results

DEMOGRAPHIC AND CLINICAL CHARACTERISTICS OF RECIPIENTS AND DONORS:

Twenty liver transplant recipients were selected for proteomics analysis, with demographic and clinical data summarized in Table 1. The primary diagnoses included liver failure (most common), hepatocellular carcinoma, and 1 case of primary biliary cirrhosis. Preoperative mean Model for End-Stage Liver Disease (MELD) scores were 26.8±12.8 for the PNF group, 14.8±7.7 for the EAD group, 16.6±13.1 for the OG group, and 17.4±7.1 for the Control group, with no significant differences observed between groups (F=0.685, P>0.05), indicating similar baseline liver function.

Clinical characteristics such as cold ischemia time (CIT) and warm ischemia time (WIT) are presented in Table 1. Statistical analysis showed no significant differences in these parameters (P>0.05), suggesting a consistent surgical environment. Seventy-five percent of the donors were male, with traumatic brain injury (60%) and cardiovascular events (30%) being the leading causes of donor death.

EXPLORATION OF PROTEOMICS IN LIVER TRANSPLANTATION AND UNVEILING KEY PROTEINS IN IRI:

iTRAQ-based quantitative proteomics was performed to analyze proteomic changes in donor livers before and after reperfusion. A total of 1 719 726 spectra were obtained, leading to the identification of 75 665 unique peptides and 6505 proteins using Mascot software (Figure 2). Of these, 6048 proteins were annotated via Gene Ontology (GO), allowing functional categorization.

Among the identified proteins, 5406 (including 33 unique proteins) were classified under biological processes, with enriched GO terms including single-organism processes, metabolic processes, cellular processes, and biological regulation. Additionally, 5417 proteins (including 91 unique) were associated with molecular functions, primarily catalytic activity and binding. Cellular component annotation identified 5783 proteins associated with cellular structures such as cells, organelles, membranes, and cell parts. A total of 4892 proteins were mapped to all 3 GO categories (biological process, molecular function, and cellular component), providing a comprehensive overview of protein functions (Figure 3A).

KEGG pathway enrichment analysis revealed that 4756 proteins (78.64% of identified proteins) were mapped to 324 pathways (Figure 3B). The most-enriched pathways included metabolic pathways (15.11%), endocytosis (3.6%), PI3K-Akt signaling (3.05%), phagosomes (2.64%), and protein processing in the endoplasmic reticulum (2.43%). Key regulatory pathways identified included PI3K-Akt (182 proteins), MAPK (107 proteins), Ras (98 proteins), mTOR (84 proteins), and apoptosis (82 proteins), all critical in LT. Comparative proteomics analysis revealed 579 differentially expressed proteins with significant upregulation or downregulation following liver reperfusion (Figure 3C).

POTENTIAL MECHANISMS OF METABOLIC REPROGRAMMING AND INFLAMMATORY RESPONSE IN DIFFERENTIALLY EXPRESSED PROTEINS OF LIVER TRANSPLANTATION:

To identify proteins associated with IRI and to evaluate their stage-specific relevance, we compared the proteomic profiles of the OG2, EAD2, and PNF2 groups against the Control group at T2. Differentially abundant proteins (DAPs) were identified using a fold-change threshold >1.2 and P<0.05. A total of 220 DAPs were identified in the PNF2 vs Control comparison, comprising 91 upregulated and 129 downregulated proteins (Figure 4A). Notably, among these, 160 proteins including 67 upregulated and 93 downregulated proteins were uniquely altered in the PNF2 group and were not significantly changed in the EAD2 or OG2 groups at the same post-reperfusion time point. This observation suggests that these 160 proteins possess specificity for the PNF outcome rather than merely reflecting a generalized response to the transplant procedure or reperfusion itself.

The top 20 enriched pathways for each group (PNF2, EAD2, OG2 vs Control) were analyzed and summarized in Table 2, with overlap visualized in a Venn diagram (Figure 4B). Two pathways – metabolic pathways and Glyoxylate and Dicarboxylate Metabolism – were co-enriched across multiple groups. Eleven pathways were specifically enriched in the PNF group, including Glycolysis/Gluconeogenesis, Fatty Acid Degradation, Tyrosine Metabolism, and others, suggesting significant metabolic reprogramming in PNF following ischemia-reperfusion injury. Complete enrichment results, including gene ratios, fold enrichment values, and P values for all significant pathways in each group, are provided in Table 3.

GO enrichment analysis showed that 91 proteins were involved in biological processes such as Glycolysis, Citrate Cycle, Tyrosine Metabolism, Fatty Acid Degradation, and others (Figure 4C). These proteins were mostly associated with cell parts and organelles in cellular components, and binding and catalytic activity in molecular function, indicating their key roles in enzymatic and regulatory processes. Results of comprehensive GO enrichment analyses for each experimental group compared to the Control group are presented in Figure 5.

The identified DAPs were enriched in several important signaling pathways, including Metabolic Pathways, IL-17 Signaling, Ras Signaling, TNF Signaling, PPAR Signaling, and Estrogen Signaling, all of which are linked to inflammation, metabolism, and immune regulation.

PNF-ASSOCIATED PROTEINS AND KEY ROLES IN OXIDATIVE STRESS AND ENERGY METABOLISM:

To validate the candidate proteins identified through proteomic analysis, MRM was performed to quantify the differential expression of target proteins in the liver. Initially, 119 proteins were selected for MRM verification based on their differential expression in IRI. Of these, 103 proteins were suitable for MRM analysis. After further screening, 40 proteins with a high fold change and clear association with hepatic injury were chosen for final MRM validation. The MRM quantitative analysis showed a significant positive correlation with the iTRAQ results (Figure 6, R=0.6465), confirming the technical consistency and reliability of the proteomic data across the 2 distinct mass spectrometry platforms.

In the comparison of the 160 uniquely differentially expressed proteins in the PNF2 and Control groups, 10 proteins were selected for further MRM validation. These proteins included: PRDX1, HGD, THIO, 6PGD, HPPD, VWF, TSP1, PCCB, GATM, and SDHB. As shown in Figure 7A, the expression levels of PRDX1, HGD, THIO, 6PGD, and HPPD were significantly downregulated in the PNF2 group compared to the Control group. Conversely, VWF was significantly upregulated (Figure 7A, 7B). The remaining 4 proteins (PCCB, TSP1, GATM, and SDHB) showed no significant differences in expression between the PNF2 and Control groups after IRI (Figure 7A, 7B). Quantitative data, including iTRAQ fold changes, MRM fold changes, MRM log2fold change, and MRM FDR values, are provided in Table 4. Proteins that were verified by MRM were further grouped according to their subcellular localization and functional roles, as shown in Table 5. Notably, all the significantly downregulated proteins were linked to oxidative stress and energy metabolism pathways. In contrast, the significantly upregulated protein VWF is primarily involved in cell adhesion and blood coagulation, and it is generally derived from the extracellular matrix. These results suggest that the identified biomarker panel reflects the generic insult of ischemia-reperfusion and PNF-specific failure of protective mechanisms during the early reperfusion phase, thereby exhibiting stage-specific and disease-specific characteristics.

Discussion

PNF is the most severe form of liver IRI, inevitably resulting in graft failure and necessitating emergency retransplantation, which carries a remarkably high mortality rate [20–22]. The multifactorial nature of PNF – often linked to extended criteria donors (ECDs), prolonged ischemia, and severe steatosis – makes early clinical prediction highly challenging. In this study, we used a rigorous iTRAQ and MRM proteomic pipeline to identify a highly specific biomarker panel (PRDX1, HGD, THIO, 6PGD, HPPD, and VWF). Crucially, the altered expression of these 6 proteins was observed at the post-reperfusion stage (T2) in grafts that subsequently developed PNF. This underscores that our findings reflect a PNF-specific molecular collapse rather than a generalized surgical or IRI response.

Rather than simply mirroring general hepatic IRI, our proteomic data reveal that PNF is characterized by a profound failure in specific metabolic and antioxidant defenses. We observed a significant downregulation of HGD and HPPD, 2 critical enzymes in phenylalanine and tyrosine catabolism [23,24]. Because these amino acids are essential glucogenic and ketogenic energy sources, the suppression of HGD and HPPD indicates a severe disruption in energy supply during the critical early reperfusion phase. Beyond mere energy deprivation, the metabolic bottleneck caused by HGD and HPPD downregulation can lead to toxic accumulation of intermediate metabolites, such as maleylacetoacetate and fumarylacetoacetate [25]. These accumulated intermediates are known to be highly reactive and hepatotoxic, directly inducing hepatocellular apoptosis and further exacerbating cellular stress [26–28]. This metabolic paralysis prevents the graft from recovering basic cellular functions, directly contributing to the irreversible nature of PNF [29].

Furthermore, the significant downregulation of THIO, 6PGD, and PRDX1 highlights a catastrophic breakdown of the graft’s reactive oxygen species (ROS) scavenging systems [30–32]. It is particularly noteworthy that 6PGD is a rate-limiting enzyme in the pentose phosphate pathway, which is the primary cellular source of NADPH. Because both the thioredoxin (THIO) and glutathione antioxidant systems strictly rely on NADPH to regenerate their active forms, the simultaneous drop in 6PGD and THIO represents a systemic collapse of the entire NADPH-dependent antioxidant axis, rather than just isolated enzyme failures. This specific depletion leaves the PNF graft defenseless against massive ROS bombardment from activated Kupffer cells upon reperfusion. Specifically, the loss of PRDX1, which normally modulates the microenvironment by inhibiting macrophage adhesion, likely exacerbates the unchecked inflammatory cascade unique to PNF [33,34].

In stark contrast to the downregulated protective enzymes, Von Willebrand Factor (VWF) was significantly upregulated in the PNF group. VWF is a major driver of platelet aggregation and microthrombus formation and thrombus formation is one of the key contributors to PNF [35]. Mechanistically, we hypothesize that the systemic collapse of antioxidant defenses (THIO/6PGD/PRDX1) leads to overwhelming oxidative damage to the liver sinusoidal endothelial cells. This severe endothelial injury triggers massive exocytosis of ultra-large VWF multimers into the microcirculation. The resulting microthrombi block blood flow, abruptly worsening local hypoxia, which in turn accelerates the metabolic failure (HGD/HPPD) and generates even more ROS. The simultaneous alteration of these 6 proteins essentially maps out a lethal, self-amplifying “vicious cycle” – metabolic failure, unbuffered oxidative stress, and microvascular thrombosis – that drives the irreversibility of PNF [36,37].

Despite these promising findings, several limitations of our study must be acknowledged. First, the definition and selection of our Control group present challenges for direct clinical applicability. We used pre-perfusion biopsies from optimal organ donors to establish a pristine molecular baseline. However, clinically, ECDs inherently possess different baseline molecular profiles than optimal donors. Second, due to the relatively small sample size and the highly restrictive criteria used to define PNF, our results must be strictly interpreted as hypothesis-generating rather than definitive. While the identified 6-protein panel is highly promising and biologically consistent with PNF pathology, it currently serves as a foundation for future targeted studies.

Conclusions

Our study used an integrated iTRAQ and MRM-based proteomic approach to identify a stage-specific biomarker panel associated with PNF after LT. By focusing on the critical post-reperfusion phase, we identified 6 proteins – PRDX1, HGD, THIO, 6PGD, HPPD (downregulated), and VWF (upregulated) – that are uniquely dysregulated in PNF grafts, indicating a correlation between this biomarker panel and the occurrence of PNF. This panel holds promise for the early identification of patients at high risk of reperfusion injury following LT, who may consequently benefit from improved and individualized therapeutic strategies. Further prospective studies are required to determine whether biomarker-guided strategies can improve outcomes in LT.